PyQA is Blackcat Informatics Inc.'s opinionated quality-assurance harness. It ships as the pyqa_lint package (install via pip install pyqa_lint) and can also be vendored as a git submodule so teams can run the Typer-based CLI (pyqa) with zero external dependencies.

TL;DR: Deterministic lint orchestration, SOLID-aware reporting, curated tool catalogs, and provenance-backed releases for modern Python repositories.

• Getting Started
• CLI Overview
• Feature Highlights
• Architecture & Docs
• Releases & Provenance
• Configuration & Tooling Schema
• Advanced Capabilities
• Contributing & Support

1. Why PyQA?
2. Installation Options
3. Getting Started
4. CLI Overview
5. Feature Highlights
6. Architecture & Docs
7. Releases & Provenance
8. Configuration & Tooling Schema
9. Advanced Capabilities
10. Contributing & Support
11. License

• Curated multi-language linting powered by ruff, pylint, bandit, mypy, pyright, sqlfluff, yamllint, and more.
• Polyglot coverage across Python, JavaScript/TypeScript, Go, Rust, SQL, YAML, TOML, Markdown, Dockerfiles, Bash, and Lua via the built-in catalog.
• Deterministic environments built with uv, with optional system-tool reuse when explicitly allowed.
• Explainable orchestration via --explain-tools and --explain-tools-json, so contributors see why tools ran or skipped.
• SOLID advice + refactor navigator driven by tree-sitter and spaCy to highlight structural issues, hotspots, and symbols.
• First-class reporting to Markdown, SARIF 2.1.0, JSON, and PR summaries, plus machine-readable tool catalogs.
• Interface-first design: runtime modules consume Protocols from pyqa.interfaces.*, keeping extensions pluggable and testable.

PyQA's catalog bundles analyzers, formatters, and value-type rules for multiple ecosystems. Highlights include:

• Python – ruff, pylint, bandit, mypy, pyright, pytest helpers, value-type analysis.
• JavaScript / TypeScript – eslint, prettier, tsc, markdownlint cross-checks for front-end code.
• Go – golangci-lint, gofmt, module tidy helpers.
• Rust – clippy, cargo fmt, cargo check integrations.
• Shell & Containers – shellcheck, shfmt, hadolint, dockerfilelint.
• Data & Config – sqlfluff (SQL), yamllint (YAML), taplo (TOML), markdownlint/mdformat (Markdown), terraformer rules.
• Scripting & DSLs – luacheck (Lua), phpcs (PHP), generic tree-sitter hooks for custom DSLs.

Refer to tooling/catalog/ for the authoritative list and metadata for each tool.

pip install pyqa_lint
# exposes the `pyqa` CLI globally

git submodule add https://github.com/paudley/pyqa_lint.git pyqa-lint

This layout keeps the managed scripts (lint, check-quality, security-scan, install, install-hooks, etc.) inside your repository without publishing a wheel.

Clone the repo and sync the toolchain:

uv sync --group dev

The shims in the repo root (./lint, ./check-quality, ./install, ./security-scan) automatically target the synced .venv.

1. git submodule add https://github.com/paudley/pyqa_lint.git pyqa-lint
2. cd pyqa-lint && uv sync --group dev
3. ./install-hooks to wire pre-commit, pre-push, and commit-msg
4. ./lint (or pyqa lint) to run the curated pipeline

Hooks and CI both execute pyqa check-quality + pyqa lint, so the same guardrails apply everywhere.

Run pyqa --help or pyqa <command> --help for detailed options.

• Explainable selection: pyqa lint --explain-tools [--json] renders the tool order, reasons, eligibility, and registry descriptions. Tests assert table + JSON output so UX changes stay stable.
• SOLID advice + highlighting: --advice adds the SOLID panel, Refactor Navigator stats, and tree-sitter/spaCy highlighting. Exporters can embed the same data via pyqa.reporting.emitters.write_pr_summary(include_advice=True).
• Value-type analysis: opt into [tool.pyqa.generic_value_types] in pyproject.toml and run pyqa lint --check-value-types-general to enforce structural traits (for example repositories must implement __len__ + __contains__).
• Deterministic envs: uv manages every dependency. PyQA reuses system binaries only when they are newer and you have not forced --use-local-linters.
• Rich reporting: emit Markdown, SARIF, JSON, or PR summaries, and pipe catalogs into ref_docs/tool-schema.json with pyqa config export-tools.
• Typed catalog + DI seams: runtime modules only import from pyqa.interfaces.*, while the reusable schema lives in src/tooling_spec/ for downstream consumers.

The REORG initiative is complete and all guidance now lives in first-class docs:

• docs/ARCHITECTURE.md – package responsibilities, SOLID guardrails, and interface boundaries.
• SELECTION.md, SOLID_CLI.md, docs/orchestration/* – orchestration internals, DI seams, progress rendering, and state machines.
• docs/tooling/ – catalog schemas, shared knobs, and authoring guides.
• docs/release/0.3.0.md – highlights for the rename + provenance work.

Key takeaways:

• pyqa.cache handles persistence and cache repositories behind pyqa.interfaces.cache.
• pyqa.analysis provides spaCy/tree-sitter integrations and refactor analytics.
• pyqa.reporting owns console/Markdown/SARIF emitters and the SOLID advice builder.
• pyqa.cli hosts the Typer app, option builders, and launcher shims (mirrored by ./lint, ./install, etc.).
• pyqa.tools defines built-in tool registrations plus parser Protocols.
• tooling_spec is a standalone distribution exporting the catalog schema for external tooling.

• Package name: pyqa_lint; CLI/import namespace: pyqa.
• v* tags trigger .github/workflows/release.yml, which:
  1. Runs uv sync --group dev
  2. Builds sdist + wheel via python -m build
  3. Uploads artefacts for inspection
  4. Generates a GitHub attestation (actions/attest-build-provenance@v1)
  5. Publishes to PyPI via pypa/gh-action-pypi-publish with attestations enabled
• Required repo secret: PYPI_API_TOKEN (trusted publisher token for pyqa_lint).
• Release checklist: bump pyproject.toml, update docs/release/<version>.md, regenerate ref_docs/tool-schema.json (uv run pyqa config export-tools --check ref_docs/tool-schema.json), commit, tag vX.Y.Z, push tag.

• Layered configuration: defaults → ~/.pyqa_lint.toml → [tool.pyqa] in pyproject.toml → <project>/.pyqa_lint.toml. Inspect with pyqa config show --root <path> --trace or compare layers via pyqa config diff.
• Strict validation: pass --strict-config (or pyqa config validate --strict) to fail on unknown keys.
• License policy: [tool.pyqa.license] declares SPDX identifiers, notices, year ranges, and exceptions enforced by pyqa check-quality.
• Quality defaults: [tool.pyqa.quality] centralizes enabled checks, schema targets, skip globs, and protected branches for hooks/CI.
• Shared knobs: [complexity] and [strictness] propagate limits to ruff, pylint, luacheck, mypy, pyright, tsc, etc. Sensitivity presets (--sensitivity low|medium|high|maximum) shift multiple thresholds together.
• Tool overrides: configure [tool.pyqa.<tool>] (or .pyqa_lint.toml's [tools.<tool>]) to customize args/env per tool; CLI flags still win for ad-hoc overrides.
• Schema exports: pyqa config schema --format markdown documents every option; pyqa config schema --format json-tools --out ref_docs/tool-schema.json and pyqa config export-tools keep the checked-in schema current.

[tool.pyqa.generic_value_types]
enabled = true

[[tool.pyqa.generic_value_types.rules]]
pattern = "myapp.repositories.*"
traits = ["iterable", "value"]
require = ["__len__", "__contains__"]
recommend = ["__repr__"]

[[tool.pyqa.generic_value_types.implications]]
trigger = "method:__len__"
require = ["__bool__"]
severity = "warning"

Run pyqa lint --check-value-types-general (or ./lint --check-value-types-general). Diagnostics such as generic-value-types:missing-required demand a suppression_valid justification when they cannot be satisfied.

./lint --validate-schema
# or
uv run python -m jsonschema -F errors \
  -i tooling/catalog/languages/python/bandit.json \
  tooling/schema/tool_definition.schema.json

The catalog schemas are intentionally reusable so other orchestrators can adopt the same contracts. See tooling/schema/SCHEMA.md and tooling/catalog/docs/SHARED_KNOBS.md for field-by-field detail.

Tree-sitter grammars are pinned in pyproject.toml and installed via uv. Install spaCy's English model once per environment:

uv run python -m spacy download en_core_web_sm

Override with PYQA_NLP_MODEL=<model> or disable highlighting entirely with --no-color in strict CI logs.

• Review CONTRIBUTING.md and the Code of Conduct before submitting PRs.
• Run uv run pyqa check-quality and uv run pytest locally prior to opening a PR.
• File issues or feature requests directly on GitHub; include the output from pyqa lint --explain-tools when reporting orchestration bugs.

This project is licensed under the MIT License. See LICENSE for details.